An explosive volume of video is available within an ‘on-demand’ context. In an on-demand system, any piece of video can be viewed by any user at any point of time. Unlike a linear or sub-linear system where some centralized scheduling is used and the users are somewhat limited in their viewing choices at a given point in time, an on-demand system makes no guess as to what a given user will want to watch when. A number of challenges exist in facilitating such a system—there is a considerable bandwidth requirement if the system is to be distributed in any way and storage technologies that are able to stream large volumes of data, perhaps in parallel are required, for example. A further challenge that has become apparent only as these systems get larger is that of navigation. Put simply, when the average consumer's video experience was limited to fewer than a hundred explicit linear channels, a simple Programming Guide and channel selector buttons on a remote control may have provided sufficient user interface to the content available. As the corpus of available content gets larger, however, fixed channel and other hierarchical ‘choice’ systems become increasingly unwieldy to operate and free-form, textual search is used instead.
When free-form search is used over a large corpus, however, it will still often return voluminous amounts of possible search results. A user needs to quickly and efficiently sort through these many options (sometimes numbering in their hundreds) in order to choose a video segment to actually watch. This search problem exists outside the video context and has been solved in various ways—textual search engines (e.g. www.Google.com, www.altavista.com, etc) already make use of so-called ‘contextual summarization’ that displays a portion of the text of the matching textual document so that a user is able to quickly assess not just that a given document matched their search but also why and how that document matched their search.